Original Article Proc IMechE Part D: J Automobile Engineering 1–15 Ó IMechE 2020 Article reuse guidelines: sagepub.com/journals-permissions DOI: 10.1177/0954407020931686 journals.sagepub.com/home/pid Semi-physical NOx and soot model for CI engines: Study of its calibration procedure and portability Hassan Karaky, Pierre Marty , Xavier Tauzia, Alain Maiboom and Gilles Mauviot Abstract A series of papers has previously presented a semi-physical model for NOx and soot emissions prediction for diesel engines. In this paper, the work is continued with an original analysis of the model’s capacity to be ported to a new engine and a sensitivity analysis of the number of training points required to obtain the desired accuracy. These two aspects are rarely developed in similar studies. Keywords Diesel engine, exhaust emissions, semi-physical NOx-soot model, model calibration procedure, model portability Date received: 18 November 2019; accepted: 23 April 2020 Introduction Diesel engines have been strongly impacted by recent emission scandals. 1 Plummeting car sales and end of production announcements by manufacturers could make one believe that this technology is doomed. However, diesel is still essential for many heavy duty applications (ships, trains, trucks, electricity generation, etc.) for which spark ignited engine’s lower efficiency cannot compete. 2 For cars, diesel engines can still com- pete, even with hybrid ones, for certain type of opera- tional profiles (rural, long range, high speed). 3 Remain the health impacts. If the proof of diesel emissions car- cinogenicity seems to be solid, 4 the death toll attributed to diesel by some studies is very controversial and pri- marily highlights the necessity to replace old cars. 5 As a matter of fact, engine manufacturers have developed numerous technologies to reduce noxious emissions (exhaust gas recirculation (EGR), high pressure injec- tion systems, sophisticated boosting systems, after- treatment devices, etc.). As a consequence, diesel engines are getting more complex making engine simu- lation even more a necessity. Trade-off between predict- ability (pure physical models) and simplicity (empirical models) is at the centre of model development reflec- tion. However, emission scandals as well as new world- wide harmonized light vehicle test procedure (WLTP) and real driving emissions (RDE) cycle testing have shed light on ‘real life’ emissions and hence on the necessity to predict correctly emissions over realistic driving cycle and no longer over theoretical ones. This trade off has become even more stringent, increasing the need for predictive (not only steady points) and fast (over a complete driving cycle) models. An hybrid approach has already been presented. 6 The combustion process is modelled via a high frequency combustion model (Barba’s approach) 7 while a thermodynamic two-zone calculation is used to evaluate combustion related variables such as adiabatic flame temperature. A semi-physical sub-model calculates in turn soot 8 and NOx 9 emissions at each exhaust valve opening. This rather original approach has been validated on a 1.6 litre Euro 5 diesel engine and has shown very good cor- relation coefficients relative to literature soot and NOx models. In this paper, the presentation of the model is pushed even further, with a special focus on an rarely discussed yet important aspect of model development, which is the study of first the ‘portability’ of the model and sec- ond its sensitivity to training point number: LHEEA (UMR CNRS 6598), E ´ cole Centrale de Nantes, Nantes, France Corresponding author: Pierre Marty, LHEEA (UMR CNRS 6598), E ´ cole Centrale de Nantes, 1, rue de la Noe ¨, 44321 Nantes, France. Email: pierre.marty@ec-nantes.fr